Bar Management Software Is Solving the Wrong Problem

What the category actually optimises for

Search bar management software and you get five kinds of product wearing one label. It is worth separating them, because they are not competing with each other and a buyer comparing them side by side is comparing answers to different questions.

• Back office and accounting suites. Close the books, automate accounts payable, cost recipes. The buyer is finance or ownership.
• Bar inventory apps. Count what is on the shelf, calculate pour cost, generate orders. The buyer is the bar manager.
• POS native reporting. Report what sold and what it cost. The buyer already bought it with the register.
• Labor and scheduling. Build the rota, forecast labor, manage tips.
• Procurement marketplaces. Get product through the door.

Every one of those is a real job and several of these products are genuinely good at theirs. Lay them end to end, though, and notice what none of them does. Not one assigns a piece of work to a person, with a deadline, and then records whether it happened.

Why did measurement get all the attention?

Not because vendors are lazy. Because measurement is the tractable problem.

Counting bottles is a well defined engineering task with a checkable answer. You can photograph a shelf, weigh a bottle, or read a flow meter, and be right or wrong in a way everyone agrees on. Building a product that reliably changes what a bartender does on a Friday night is a much harder problem involving people, and it does not demo as cleanly.

So the category optimised the measurable half, and got good at it. Counts that took a Sunday now take fifteen minutes. Invoices that were keyed by hand are read automatically. Those are real wins and nobody should give them back.

But the constraint moved. When counting took six hours, faster counting was the whole game. Now that it takes fifteen minutes, the bottleneck is downstream, and the category has not followed it there.

The four causes problem

Here is the concrete version, and it is why we keep saying attribution rather than measurement.

Your variance says you are down roughly four percent on a well spirit. That single number is consistent with at least four different situations:

• Overpouring. One or two people, probably on specific shifts, pouring heavy. The fix is coaching and possibly a jigger policy.
• Unentered comps. Drinks given away without being rung. The fix is a process change and a conversation.
• A short delivery. Signed for at the door without being checked. The fix is receiving discipline, and possibly a credit from the supplier.
• Breakage nobody logged. The fix is a logging habit, and the money is already gone.

Those have completely different remedies, different owners and different costs. Acting on the wrong one is worse than doing nothing, because you spend political capital coaching somebody about a pour when the problem was the loading bay.

No amount of improvement in counting accuracy separates these four. The separating information is in the transaction stream: who rang what, when, in what pattern, against what the rest of the team did. That is a different data source and a different product.

There is evidence for how much that stream holds. Washington University in St. Louis reported in 2013 that when 392 restaurants installed monitoring that reads point of sale data, theft fell 22 percent and revenue per restaurant rose about 7 percent, and the published study (Management Science, 2015) traced the gain to changed behaviour rather than to staff turnover.

What should the buying question be?

If you are evaluating tools in this category, the checkbox comparison will mislead you, because an all in one suite wins a feature count by construction and still may not do the thing you need.

A better test: when this software has finished with me, what am I holding? A number, a report and a dashboard are all the same answer. They are the input to the work. If nothing in the product produces a named piece of work with an owner and a due date, then that step is still your job, and you should price the tool accordingly.

We applied that test honestly to nine named products, including several we think are very good, in the comparisons. Each one says plainly where the other product is the better fit, because a comparison that does not is worth nothing to somebody actually deciding.

The part we are not claiming

Execution software does not remove judgement. Deciding whether a void pattern is theft, a training gap or a badly built menu button needs context no system has, and anyone selling you automated certainty about that is overselling.

What it removes is the labour around the judgement: finding the pattern early, attaching the evidence, getting it to the right person, ranking it against everything else, and remembering what was decided. That is most of the work and almost none of the thinking.

The longer version of this argument is at the data to action gap, and the product version is at why CoreTAP.